基于近红外光谱的芒果采后品质与贮放潜力预判的无损检测模型

可溶性固形物含量（soluble solids content, SSC）和pH是决定芒果内在品质的关键因素,贮放潜力是果商进行芒果销售决策时首要的参考指标。本研究以海南省三亚市代表性的芒果品种‘台农’为材料,利用NIRMagic2400型近红外光谱仪,连续采集果实从采摘到完熟过程中在600~1100 nm波长的近红外吸收光谱,以经典方法实测果实SSC和pH,建立芒果采摘后SSC、pH变化和贮放潜力预判的无损检测模型。结果表明：在600~670 nm的波长范围内,采摘后未后熟的芒果对近红外光的吸光度随波长的增加而增加,并在670 nm达到峰值,随后吸光度快速降低,在725 nm左右达到谷值;采摘后达到完熟的芒果在600-700 nm波长范围内吸光度持续下降,并在700 nm处达到谷值。受果皮颜色差异等影响,不同芒果个体在704~746 nm区域的吸光度出现较大的分离,之后在725~1025 nm整体呈缓慢上升的趋势,在1025 nm左右达到第二个峰值。实测结果显示SSC在芒果采摘后0~5 d快速增,第6和第7天变化较小,期间的前4 d的pH保持稳定增加,之后迅速提升。使用Kennard-Stone算法将芒果样本的SSC和pH实测数据划分为校正集和预测集,测试多元散射校正、标准正态变换、SG卷积导数、SG卷积平滑等9种对近红外光谱数据进行预处理的方法,发现矢量归一化最适合SSC光谱数据的处理,多元散射校正最适合pH光谱数据的预处理,建立的SSC和pH的最佳偏最小二乘法（PLS）模型的校正相关系数分别为0.952和0.936,校正均方根误差分别为1.055和0.184,预测相关系数分别为0.959和0.918,预测均方根误差分别为0.974和0.202;采用偏最小二乘法建立的芒果贮放潜力预判模型的正确率为96.9%。以上结果表明,基于近红外光谱所建立的芒果无损检测模型能够较可靠地检测芒果采摘后的SSC、pH动态变化及贮放潜力。研究结果对提升基于内在品质的芒果分级与选品能力,预测芒果的最佳销售时间及选择销售市场等都具有重要意义。

Near-Infrared Spectroscopy Nondestructive Testing Model for Mango Fruit Quality Assay and Storage Potential Prediction

Soluble solids content (SSC) and pH are key factors determining the intrinsic quality of mango. In this study, ‘Tai Nong’, a representative mango cultivar of Sanya, Hainan, China, was tested. The near-infrared spectrum (600- 1100 nm) was collected every day from mango harvest to the full ripening using a NIRMagic 2400 near-infrared spectrometer. The actual soluble solids content and pH change was assayed by classic destructive methods to validate the non-destructive testing model. The results showed that under diffuse transmission detection mode, the near-infrared light absorbance of pre-fully-ripe mango increased from 600 nm to 670 nm, peaked at 670 nm, then the absorbance decreased rapidly and reached the valley value at around 725 nm. While the absorbance of the fully-ripe mangoes exhibited continuous decline in the wavelength range of 600-700 nm, and reached the lowest value at 700 nm. Due to peel color differences, individual mangoes presented larger absorbance separation between 704 nm to 746 nm. Slow absorbance increase was recorded from 725 nm to 1025 nm, and reached the second peak at around 1025 nm. Destructive tests showed that mango soluble solids content increased rapidly on the 5th day after the harvest, changed slightly on the 6th and 7th day, whereas the fruit pH was almost stable in the first 4 days, then quickly increased. The measured samples were then divided into the correction and prediction sets using the Kennard-Stone algorithm. Preprocessing methods including multiplicative scatter correction (MSC), standard normal variate transform (SNV), SG convolution derivative, SG convolution smoothing and different combinations were tested. It was revealed that the vector normalization and the multiplicative scatter correction were the best for soluble solids content and pH spectral data processing, respectively. The correlation coefficients (R_c) of the soluble solids content and pH was 0.952 and 0.936 respectively, when the least partial square method (PLS) model was used. The root mean square error of correction (RMSEC) was 1.055 and 0.184. The predicted correlation coefficients (R_p) of the soluble solids content and pH was 0.959 and 0.918 using the least partial square method model, whereas the root mean square error of prediction (RMSEP) was 0.974 and 0.202, respectively. The mango storage potential prediction accuracy rate was 96.4% when the model was established using the least partial square method. The set of data confirmed the establishment of a reliable nondestructive testing model for mango fruit soluble solids content and pH assay and storage potential prediction. Our results are useful in mango harvest determination, fruit grading, and predicting the optimal time for selling based on intrinsic fruit quality.